Biologically Plausible Neural Network Framework for Continual Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Artificial neural networks (ANNs) suffer from catastrophic forgetting, forgetting previously learned information when learning new tasks, due to their inability to balance stability and plasticity, unlike biological neural networks which adapt to changing environments.
Innovation Solution
A biologically plausible framework for artificial neural networks is introduced, featuring exclusively excitatory and inhibitory neurons, dendritic segments, and mechanisms like Hebbian updates and synaptic consolidation, along with experience replay, to mimic biological learning processes and prevent forgetting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If standard artificial neural networks are trained sequentially on new tasks, then the network learns new information efficiently, but it completely forgets previously learned information (catastrophic forgetting)
Solution Approach 1:
The patent applies preliminary action by storing task prototypes (compressed representations of previous tasks) in advance before new tasks are learned. These prototypes are saved in a memory buffer and replayed during subsequent training, allowing the network to review and retain previous knowledge before adapting to new tasks, thus preventing catastrophic forgetting while maintaining learning efficiency.
2Loss of information
If the neural network retains all previous information, then knowledge is preserved, but the network cannot adapt to new tasks efficiently
Solution Approach 1:
The patent extracts only the essential features of previous tasks by compressing them into task prototypes (e.g., using autoencoders or summary statistics). Instead of retaining all raw data from previous tasks, the network stores compressed representations that capture the most important patterns. This extraction allows efficient storage and replay of previous knowledge without interfering with new task learning.
3Productivity
If the network uses standard learning mechanisms, then training is computationally efficient, but the learning process causes abrupt erasure of previous task performance
Solution Approach 1:
The patent implements feedback by replaying stored task prototypes during the training process on new tasks. The network periodically reviews previous task representations and adjusts its weights to maintain performance on both old and new tasks. This feedback mechanism ensures that learning remains efficient while preventing the abrupt erasure of previous knowledge that occurs in standard learning mechanisms.
Data Source
AI summary
A computer-implemented method for general continual learning (CL) in artificial neural network that provides a biologically plausible framework for continual learning which incorporates different mechanisms inspired by the brain. The underlying model comprises separate populations of exclusively excitatory and exclusively inhibitory neurons in each layer which adheres to Dale's principle and the excitatory neurons (mimicking pyramidal cells) are augmented with dendrite-like structures for context-dependent processing of information. The dendritic segments process an additional context signal encoding task information and subsequently modulate the feedforward activity of the excitatory neuron. Additionally, it provides an efficient mechanism for controlling the sparsity in activations using k-WTA (k-Winners-Take-All) activations and Heterogeneous dropout mechanism that encourages the model to use a different set of neurons for each task. This provides an effective approach for maintaining a balance between reusability of features and interference which is critical for enabling CL. Furthermore, it complements the error-based learning with the “fire together, wire together” learning paradigm which further strengthen the association between the context signal and dendritic segments which process them and facilitates context-dependent gating. To further mitigate forgetting, it incorporates synaptic consolidation in conjunction with experience replay.